Identifying patient experience from online resources via sentiment analysis and topic modelling

Positive patient experience is crucial for retaining patient loyalty and in understanding and acting upon limitations of the treatment or service provided. Online platforms, such as websites and forums, are excellent sources for collecting more reliable feedback, as they provide anonymity and ease of use to the patients. Information from online sources can be vast and unstructured, thereby making patient feedback analysis challenging. Recent advancements in text mining and sentiment analysis approaches can enable automated and granular analysis of patient feedback. In this paper, we present our research, for which we applied and evaluated text mining and machine learning models to a patient feedback database obtained from the NHS Choices website to predict the patient sentiment in the database. There were two iterations to our research. First, we applied a linguistic approach using machine learning and dictionary scoring algorithms to predict patient sentiment from patient feedback and the predicted sentiment was validated against the ratings provided by the patients in the database. Second, a topic modelling approach was applied to identify "themes" within patient feedback so as to understand better the nature of the associated sentiment score, thereby providing a richer understanding of patient opinion.